cert
cert-1
cert-2

Pass IAPP AIGP Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
AIGP Exam - Verified By Experts
AIGP Premium Bundle
$29.98

AIGP Premium Bundle

$79.99
$109.97
  • Premium File 222 Questions & Answers. Last update: Nov 15, 2026
  • Training Course 124 Video Lectures
  • Study Guide 566 Pages
 
$109.97
$79.99
accept 63 downloads in last 7 days
block-screenshots
Purchase IAPP AIGP Exam Training Products Individually
IAPP AIGP Premium File
Premium File 222 Q&A
$76.99$69.99
IAPP AIGP Video Training Course
Training Course 124 Lectures
$21.99 $19.99
IAPP AIGP PDF Study Guide
Study Guide 566 Pages
$21.99 $19.99
IAPP AIGP Practice Test 66

Question 650

A support chatbot is exposed to untrusted uploaded documents yet designers assume all retrieved text is safe to follow. Model assurance reviewers need a documented control before approving continued experimentation. Which specific implementation best addresses the identified gap?

  1. Conduct an AI impact assessment with accessible stakeholder participation and documented mitigations.
  2. Write a bounded use-case specification with human authority, exclusions and measurable patient-safety objectives.
  3. Apply a documented likelihood-severity methodology and higher safeguards for rare but severe harms.
  4. Threat-model prompt injection and data exfiltration; segregate instructions, restrict tools and validate outputs.

Explanation

Correct Answer

D:

Threat-model prompt injection and data exfiltration; segregate instructions, restrict tools and validate outputs. Correct. It directly reduces or governs the specific documented risk. Generative AI development requires threat modeling for instruction confusion and downstream actions.

Incorrect Answers

A:

Conduct an AI impact assessment with accessible stakeholder participation and documented mitigations. Not selected. That action treats another control objective, not the specific gap described. It belongs with Stakeholder impact assessment: Impact assessment should consider people and groups who bear risk, not only the organization buying AI. The present case instead concerns Generative AI misuse threat model.

B:

Write a bounded use-case specification with human authority, exclusions and measurable patient-safety objectives. Not selected. That action treats another control objective, not the specific gap described. It belongs with Use-case problem framing: A defensible AI design begins with the problem and context, not a preference for a model technology. The present case instead concerns Generative AI misuse threat model.

C:

Apply a documented likelihood-severity methodology and higher safeguards for rare but severe harms. Not selected. That action treats another control objective, not the specific gap described. It belongs with Risk severity and probability matrix: Low observed frequency in a limited test cannot automatically dismiss a severe AI hazard. The present case instead concerns Generative AI misuse threat model.

Question 651

A support chatbot is exposed to untrusted uploaded documents yet designers assume all retrieved text is safe to follow. Model assurance reviewers need a documented control before approving continued experimentation. Which acceptance test most directly verifies the relevant control?

  1. Measure performance on representative cohorts using the exact proposed production threshold.
  2. Reconstruct a model-choice decision and confirm its evidence and rationale can be audited.
  3. Verify the assessment includes relevant harms, likelihood, severity and accountable treatment actions.
  4. Challenge the system with adversarial documents and confirm no secret disclosure or unauthorized tool action.

Explanation

Correct Answer

D:

Challenge the system with adversarial documents and confirm no secret disclosure or unauthorized tool action. Correct. It yields measurable evidence that the case-specific safeguard works. Generative AI development requires threat modeling for instruction confusion and downstream actions.

Incorrect Answers

A:

Measure performance on representative cohorts using the exact proposed production threshold. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Thresholds and acceptance metrics: Governance metrics should reflect the actual harm of error types, not aggregate accuracy alone. The present case instead concerns Generative AI misuse threat model.

B:

Reconstruct a model-choice decision and confirm its evidence and rationale can be audited. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Design record and decision provenance: Design documentation makes technical tradeoffs governable and defensible across staff changes. The present case instead concerns Generative AI misuse threat model.

C:

Verify the assessment includes relevant harms, likelihood, severity and accountable treatment actions. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Stakeholder impact assessment: Impact assessment should consider people and groups who bear risk, not only the organization buying AI. The present case instead concerns Generative AI misuse threat model.

Question 652

A support chatbot is exposed to untrusted uploaded documents yet designers assume all retrieved text is safe to follow. Model assurance reviewers need a documented control before approving continued experimentation. For a defensible decision, choose TWO: one diagnosis and one controlled response.

  1. Apply a documented likelihood-severity methodology and higher safeguards for rare but severe harms.
  2. Trace trust boundaries from uploaded content to model prompts, tool calls and authorized business actions.
  3. Threat-model prompt injection and data exfiltration; segregate instructions, restrict tools and validate outputs.
  4. Identify decision costs, class imbalance, subgroup error rates and the operating threshold.

Explanation

Correct Answers

B:

Trace trust boundaries from uploaded content to model prompts, tool calls and authorized business actions. Correct. As the first, evidence-gathering half of the requested pair, this establishes the specific risk boundary. Generative AI development requires threat modeling for instruction confusion and downstream actions.

C:

Threat-model prompt injection and data exfiltration; segregate instructions, restrict tools and validate outputs. Correct. As the second, implementation half of the requested pair, this directly applies the needed safeguard. Generative AI development requires threat modeling for instruction confusion and downstream actions.

Incorrect Answers

A:

Apply a documented likelihood-severity methodology and higher safeguards for rare but severe harms. Not selected. It addresses a different governance mechanism, not the required assessment-and-control pair. It belongs with Risk severity and probability matrix: Low observed frequency in a limited test cannot automatically dismiss a severe AI hazard. The present case instead concerns Generative AI misuse threat model.

D:

Identify decision costs, class imbalance, subgroup error rates and the operating threshold. Not selected. It addresses a different governance mechanism, not the required assessment-and-control pair. It belongs with Thresholds and acceptance metrics: Governance metrics should reflect the actual harm of error types, not aggregate accuracy alone. The present case instead concerns Generative AI misuse threat model.

Question 653

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. The development team must show evidence before promoting the next model version. Before selecting a remedy, what should the accountable reviewer investigate first?

  1. Select foreseeable corruption, adversarial perturbation and distribution-shift scenarios to test.
  2. Assess explanation stability, fidelity and whether the method matches model and use-case limitations.
  3. Inspect collection origin, licenses, consent where applicable, lineage and transformation history.
  4. Measure sample coverage, label consistency, subgroup representation and production-population mismatch.

Explanation

Correct Answer

C:

Inspect collection origin, licenses, consent where applicable, lineage and transformation history. Correct. It establishes the relevant factual risk boundary before implementation. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

A:

Select foreseeable corruption, adversarial perturbation and distribution-shift scenarios to test. Not selected. That check primarily diagnoses a different governance issue. It belongs with Adversarial robustness testing: Security and robustness tests must cover plausible operating and adversarial conditions. The present case instead concerns Training data provenance.

B:

Assess explanation stability, fidelity and whether the method matches model and use-case limitations. Not selected. That check primarily diagnoses a different governance issue. It belongs with Interpretability testing: An explanation tool may be misleading unless its reliability and limitations are evaluated. The present case instead concerns Training data provenance.

D:

Measure sample coverage, label consistency, subgroup representation and production-population mismatch. Not selected. That check primarily diagnoses a different governance issue. It belongs with Data quality and representativeness: Training data quality includes representativeness for the intended use population. The present case instead concerns Training data provenance.

Question 654

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. The development team must show evidence before promoting the next model version. What action should the accountable team prioritize to govern this use responsibly?

  1. Maintain reproducible experiments and protected evaluation artifacts with traceable approvals.
  2. Accept only traceable and lawfully usable data with documented source and processing stages.
  3. Use temporal or group-aware splits and isolate evaluation records from training and tuning.
  4. Add robustness tests, input controls and fallback behavior for unsafe predictions.

Explanation

Correct Answer

B:

Accept only traceable and lawfully usable data with documented source and processing stages. Correct. It directly reduces or governs the specific documented risk. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

A:

Maintain reproducible experiments and protected evaluation artifacts with traceable approvals. Not selected. That action treats another control objective, not the specific gap described. It belongs with Testing evidence and reproducibility: Reproducibility makes AI testing evidence reviewable rather than anecdotal. The present case instead concerns Training data provenance.

C:

Use temporal or group-aware splits and isolate evaluation records from training and tuning. Not selected. That action treats another control objective, not the specific gap described. It belongs with Dataset leakage and temporal splits: A contaminated test set can give convincing but invalid evidence of generalization. The present case instead concerns Training data provenance.

D:

Add robustness tests, input controls and fallback behavior for unsafe predictions. Not selected. That action treats another control objective, not the specific gap described. It belongs with Adversarial robustness testing: Security and robustness tests must cover plausible operating and adversarial conditions. The present case instead concerns Training data provenance.

Question 655

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. The development team must show evidence before promoting the next model version. What result would provide the strongest case-specific assurance?

  1. Measure agreement and error on a reviewed gold-standard sample before training.
  2. Trace sampled training records from ingestion back to an authorized source and permitted transformation.
  3. Measure hazard-specific failure rates under glare, obstruction and attack conditions.
  4. Test similar inputs for explanation consistency and verify reviewers do not treat correlates as proven causes.

Explanation

Correct Answer

B:

Trace sampled training records from ingestion back to an authorized source and permitted transformation. Correct. It yields measurable evidence that the case-specific safeguard works. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

A:

Measure agreement and error on a reviewed gold-standard sample before training. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Label quality and inter-rater agreement: Ground-truth quality shapes model behavior and can hide systematic harm. The present case instead concerns Training data provenance.

C:

Measure hazard-specific failure rates under glare, obstruction and attack conditions. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Adversarial robustness testing: Security and robustness tests must cover plausible operating and adversarial conditions. The present case instead concerns Training data provenance.

D:

Test similar inputs for explanation consistency and verify reviewers do not treat correlates as proven causes. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Interpretability testing: An explanation tool may be misleading unless its reliability and limitations are evaluated. The present case instead concerns Training data provenance.

Question 656

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. The development team must show evidence before promoting the next model version. What should the owner do to prevent further harm from this specific failure?

  1. Discard contaminated benchmark claims and rerun controlled evaluation before release.
  2. Quarantine undocumented data and retrain or re-evaluate models if unauthorized material was used.
  3. Remove problematic derived features and invalidate affected model releases when lineage cannot be verified.
  4. Provide accessible non-AI service paths while underrepresented speech performance is corrected.

Explanation

Correct Answer

B:

Quarantine undocumented data and retrain or re-evaluate models if unauthorized material was used. Correct. It limits further reliance or exposure while the risk remains unresolved. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

A:

Discard contaminated benchmark claims and rerun controlled evaluation before release. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Dataset leakage and temporal splits: A contaminated test set can give convincing but invalid evidence of generalization. The present case instead concerns Training data provenance.

C:

Remove problematic derived features and invalidate affected model releases when lineage cannot be verified. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Data lineage for transformations: Lineage tracks how raw data becomes model-ready features, including indirect sensitive signals. The present case instead concerns Training data provenance.

D:

Provide accessible non-AI service paths while underrepresented speech performance is corrected. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Data quality and representativeness: Training data quality includes representativeness for the intended use population. The present case instead concerns Training data provenance.

Question 657

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. Model assurance reviewers need a documented control before approving continued experimentation. What should the AI governance lead examine before changing the system?

  1. Assess explanation stability, fidelity and whether the method matches model and use-case limitations.
  2. Audit labeling guidance, disagreement patterns, annotator context and ambiguous categories.
  3. Inspect collection origin, licenses, consent where applicable, lineage and transformation history.
  4. Trace raw data, feature engineering, joins and transformations to the final training inputs.

Explanation

Correct Answer

C:

Inspect collection origin, licenses, consent where applicable, lineage and transformation history. Correct. It establishes the relevant factual risk boundary before implementation. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

A:

Assess explanation stability, fidelity and whether the method matches model and use-case limitations. Not selected. That check primarily diagnoses a different governance issue. It belongs with Interpretability testing: An explanation tool may be misleading unless its reliability and limitations are evaluated. The present case instead concerns Training data provenance.

B:

Audit labeling guidance, disagreement patterns, annotator context and ambiguous categories. Not selected. That check primarily diagnoses a different governance issue. It belongs with Label quality and inter-rater agreement: Ground-truth quality shapes model behavior and can hide systematic harm. The present case instead concerns Training data provenance.

D:

Trace raw data, feature engineering, joins and transformations to the final training inputs. Not selected. That check primarily diagnoses a different governance issue. It belongs with Data lineage for transformations: Lineage tracks how raw data becomes model-ready features, including indirect sensitive signals. The present case instead concerns Training data provenance.

Question 658

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. Model assurance reviewers need a documented control before approving continued experimentation. What is the most directly applicable governance control for this situation?

  1. Apply legally appropriate sampling, fairness evaluation and mitigation with job-related validation.
  2. Validate explanation methods and document what attributions do and do not establish.
  3. Accept only traceable and lawfully usable data with documented source and processing stages.
  4. Record end-to-end feature lineage and review proxy risks in engineered attributes.

Explanation

Correct Answer

C:

Accept only traceable and lawfully usable data with documented source and processing stages. Correct. It directly reduces or governs the specific documented risk. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

A:

Apply legally appropriate sampling, fairness evaluation and mitigation with job-related validation. Not selected. That action treats another control objective, not the specific gap described. It belongs with Bias measurement and mitigation: Historic training labels can reproduce inequity even when protected fields are omitted. The present case instead concerns Training data provenance.

B:

Validate explanation methods and document what attributions do and do not establish. Not selected. That action treats another control objective, not the specific gap described. It belongs with Interpretability testing: An explanation tool may be misleading unless its reliability and limitations are evaluated. The present case instead concerns Training data provenance.

D:

Record end-to-end feature lineage and review proxy risks in engineered attributes. Not selected. That action treats another control objective, not the specific gap described. It belongs with Data lineage for transformations: Lineage tracks how raw data becomes model-ready features, including indirect sensitive signals. The present case instead concerns Training data provenance.

Question 659

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. Model assurance reviewers need a documented control before approving continued experimentation. What should an independent reviewer test before signing off on this specific risk?

  1. Trace sampled training records from ingestion back to an authorized source and permitted transformation.
  2. Evaluate both error differences and selection effects on representative groups after mitigation.
  3. Recreate selected features from source data and confirm sensitive transformations are visible.
  4. Measure hazard-specific failure rates under glare, obstruction and attack conditions.

Explanation

Correct Answer

A:

Trace sampled training records from ingestion back to an authorized source and permitted transformation. Correct. It yields measurable evidence that the case-specific safeguard works. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

B:

Evaluate both error differences and selection effects on representative groups after mitigation. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Bias measurement and mitigation: Historic training labels can reproduce inequity even when protected fields are omitted. The present case instead concerns Training data provenance.

C:

Recreate selected features from source data and confirm sensitive transformations are visible. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Data lineage for transformations: Lineage tracks how raw data becomes model-ready features, including indirect sensitive signals. The present case instead concerns Training data provenance.

D:

Measure hazard-specific failure rates under glare, obstruction and attack conditions. Not selected. That test would not demonstrate this particular risk has been addressed. It belongs with Adversarial robustness testing: Security and robustness tests must cover plausible operating and adversarial conditions. The present case instead concerns Training data provenance.

Question 660

A vendor offers a large image dataset with no trace of collection rights, source dates or redistribution conditions. Model assurance reviewers need a documented control before approving continued experimentation. What is the most relevant fallback when the promised safeguard cannot be demonstrated?

  1. Withdraw unverifiable performance claims and rerun controlled tests before launch.
  2. Switch to a validated safe mode when input quality falls outside the assessed operating domain.
  3. Quarantine undocumented data and retrain or re-evaluate models if unauthorized material was used.
  4. Suspend adverse automated decisions when discriminatory outcomes cannot be justified or corrected.

Explanation

Correct Answer

C:

Quarantine undocumented data and retrain or re-evaluate models if unauthorized material was used. Correct. It limits further reliance or exposure while the risk remains unresolved. Data provenance establishes what entered training and under what conditions.

Incorrect Answers

A:

Withdraw unverifiable performance claims and rerun controlled tests before launch. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Testing evidence and reproducibility: Reproducibility makes AI testing evidence reviewable rather than anecdotal. The present case instead concerns Training data provenance.

B:

Switch to a validated safe mode when input quality falls outside the assessed operating domain. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Adversarial robustness testing: Security and robustness tests must cover plausible operating and adversarial conditions. The present case instead concerns Training data provenance.

D:

Suspend adverse automated decisions when discriminatory outcomes cannot be justified or corrected. Not selected. That response contains a different failure mode and leaves the stated risk unresolved. It belongs with Bias measurement and mitigation: Historic training labels can reproduce inequity even when protected fields are omitted. The present case instead concerns Training data provenance.

Artificial Intelligence Governance Professional (AIGP) Practice Tests